Foundation Model Integration, Data Management, and Compliance
AWS AIP-C01 Domain Coverage
Content Domain 1: Foundation Model Integration, Data Management, and Compliance
This folder provides deep-dive study material for Tasks 1.1 through 1.6 using the same practical style as the other folders in this repository. The examples are grounded in the MangaAssist architecture so the user stories stay concrete instead of becoming generic certification notes.
The goal of this folder is not just to restate the exam guide. It is to help you think like a GenAI architect who must translate business requirements into safe, scalable, testable systems on AWS.
Task-to-File Mapping
| Task | File | What You Learn |
|---|---|---|
| 1.1 Analyze requirements and design GenAI solutions | 01-task-1.1-solution-design-and-requirements.md | How to convert business goals into architecture choices, POCs, and reusable standards |
| 1.2 Select and configure FMs | 02-task-1.2-foundation-model-selection-and-configuration.md | How to choose models, switch providers dynamically, design resilient fallbacks, and manage customized models |
| 1.3 Implement data validation and processing pipelines for FM consumption | 03-task-1.3-data-validation-and-processing-pipelines.md | How to validate, normalize, enrich, and format multimodal data before it reaches a model |
| 1.4 Design and implement vector store solutions | 04-task-1.4-vector-store-solutions.md | How to choose vector architectures, metadata schemes, connectors, and data-maintenance strategies |
| 1.5 Design retrieval mechanisms for FM augmentation | 05-task-1.5-retrieval-mechanisms-for-fm-augmentation.md | How to chunk content, choose embeddings, configure search, improve query handling, and standardize retrieval access |
| 1.6 Implement prompt engineering strategies and governance for FM interactions | 06-task-1.6-prompt-engineering-and-governance.md | How to design prompt frameworks, multi-turn context systems, governance, regression testing, optimization loops, and prompt flows |
| Scenario map | 07-domain-1-scenario-map.md | A domain-wide blueprint for breaking large scenario files into smaller, reviewable deep dives |
Skill Coverage
| Task | Skills Covered |
|---|---|
| 1.1 | 1.1.1 architecture design, 1.1.2 proof-of-concept validation, 1.1.3 standardized technical components |
| 1.2 | 1.2.1 model assessment and selection, 1.2.2 dynamic routing and provider switching, 1.2.3 resilience patterns, 1.2.4 customization lifecycle management |
| 1.3 | 1.3.1 data validation, 1.3.2 multimodal processing, 1.3.3 inference formatting, 1.3.4 data enhancement |
| 1.4 | 1.4.1 vector architecture, 1.4.2 metadata frameworks, 1.4.3 performance at scale, 1.4.4 integration components, 1.4.5 maintenance systems |
| 1.5 | 1.5.1 chunking, 1.5.2 embedding selection, 1.5.3 vector search deployment, 1.5.4 advanced search, 1.5.5 query handling, 1.5.6 standardized access mechanisms |
| 1.6 | 1.6.1 instruction frameworks, 1.6.2 conversational context systems, 1.6.3 prompt governance, 1.6.4 prompt QA, 1.6.5 iterative prompt optimization, 1.6.6 complex prompt systems |
How to Use This Folder
| Use Case | Recommended Path |
|---|---|
| AIP-C01 Exam Prep | Read files 01 through 06 in order because each task builds on the previous one |
| System Design Interviews | Start with 01, 02, 04, and 05 because they cover the highest-signal architecture tradeoffs |
| Bedrock Project Delivery | Start with 01, 02, and 06, then use 03 through 05 to harden the data and retrieval layers |
| RAG Platform Work | Focus on 03, 04, and 05 first because retrieval quality begins with content quality and index hygiene |
| PromptOps / LLMOps | Focus on 02 and 06, then connect back to 03 for input quality and 05 for retrieval grounding |
File Structure Convention
Each task file follows the same pattern:
- AIP-C01 Mapping - Domain, task, and scope
- Task Goal - What the task is really trying to achieve in production
- Task User Story - The business-facing story for the whole task
- Task Architecture View - A Mermaid diagram showing how the parts connect
- Per-Skill Deep Dive - A dedicated subsection for every skill under the task
- Acceptance Signals - What success looks like in a real program
- Intuition Gained After the Task - The deeper instinct you should carry into design and interview discussions
Scenario Pack Convention
The scenario material is being refactored to a cleaner study pattern:
03-scenarios-and-runbooks.mdbecomes the skill index - it explains why the scenarios matter and links to the individual deep dives.scenarios/*.mdholds one scenario per file - each file focuses on one failure mode, one design tension, and one runbook path.- Each scenario now includes questions to ask - this turns the docs into design-review and interview-prep material instead of just long incident notes.
Task 1.1 already follows this split-scenario structure. The new 07-domain-1-scenario-map.md file shows how to extend the same approach across the rest of Domain 1.
Relationship to Existing Content
These folders already contain adjacent material. This Domain 1 folder builds on them instead of duplicating them:
- Prompt-Engineering/ - Prompt patterns, hardening, and evaluation
- Fine-Tuning-Foundational-Models/ - Customization strategies such as LoRA, adapters, and tuning tradeoffs
- Model-Inference/ - Inference pipelines, endpoint choices, and model-evaluation context
- Performance-Optimization-User-Stories/ - Latency optimization for orchestration, retrieval, and response generation
- Cost-Optimization-User-Stories/ - Cost-aware model routing, caching, and RAG patterns
- Security-Privacy-Guardrails/ - Safety, abuse prevention, and compliance-oriented controls
- Evaluation-Systems-GenAI/ - How to measure the quality of what you build in Domain 1
Core Theme Across Domain 1
Domain 1 is about building the layer before evaluation and troubleshooting become possible. If you make weak choices here, the later domains become expensive band-aids:
- Weak architecture creates permanent cost and latency problems.
- Weak data pipelines create hallucinations that no prompt can fully fix.
- Weak vector-store hygiene creates retrieval noise that looks like model failure.
- Weak prompt governance turns every prompt change into an untracked production risk.
The six task files below are designed to help you build that foundation well.